Not too sure about that, that might be the case but currently, they would need much more training to not mess up facial features, to make images truly lifelike and to follow prompt instructions better.
I’ve used dalle a fair bit and I came to the conclusion that you will never get a truly accurate representation of a person, such as hair on a bald persons head, stubble turning into a moustache, tons of wrinkles for no reason, etc. It only seems good at generating cartoon characters, even then though, there are still inaccuracies.
It would be so hilarious to revisit this confident (dare I say it: AI levels of confidence!) statement in two years when AI has crashed like every other previous wave of AI tech.
That’s not true. DeepSeek showed that smart people are working on it that are not entirely motivated by just making a cheaper way to advertise. I think the wests a implementation of AI is so horribly corrupted by capitalist incentives. I think we’ll continue to see advancements out of China and Asia. It’s definitely not “just throw more data at it”.
DeepSeek showed that the Western “just throw more data at it” nonsense is a failed strategy, and certainly China is going to continue developing this technology further. Their approach is much smarter, they build different models for different problems instead of trying to build a general purpose model that can do everything at once. Then DeepSeek can switch between different models to solve different problems, instead of trying to make every problem conform to their one general purpose model.
Let me clarify - I don’t think any advancement of this technology is going to come out of Silicon Valley. They’re all trying to build the Homer Car.
Trust me, you’re going to have zero ability to discern what is AI generated in less than two years.
I already can’t easily identify most of it I’m screwed
Not too sure about that, that might be the case but currently, they would need much more training to not mess up facial features, to make images truly lifelike and to follow prompt instructions better.
I’ve used dalle a fair bit and I came to the conclusion that you will never get a truly accurate representation of a person, such as hair on a bald persons head, stubble turning into a moustache, tons of wrinkles for no reason, etc. It only seems good at generating cartoon characters, even then though, there are still inaccuracies.
It would be so hilarious to revisit this confident (dare I say it: AI levels of confidence!) statement in two years when AI has crashed like every other previous wave of AI tech.
You’re on, I’m sorry for what’s coming.
You will be when this latest AI hype turns into the next AI winter, yes.
Their only solution is “throw more data at it!”
The technology definitely can get better, but you shouldn’t assume it will. Just look at the people developing it.
That’s not true. DeepSeek showed that smart people are working on it that are not entirely motivated by just making a cheaper way to advertise. I think the wests a implementation of AI is so horribly corrupted by capitalist incentives. I think we’ll continue to see advancements out of China and Asia. It’s definitely not “just throw more data at it”.
DeepSeek showed that the Western “just throw more data at it” nonsense is a failed strategy, and certainly China is going to continue developing this technology further. Their approach is much smarter, they build different models for different problems instead of trying to build a general purpose model that can do everything at once. Then DeepSeek can switch between different models to solve different problems, instead of trying to make every problem conform to their one general purpose model.
Let me clarify - I don’t think any advancement of this technology is going to come out of Silicon Valley. They’re all trying to build the Homer Car.